RESEARCH

Control design and system identification from data.

We study how to design controllers and identify models using data collected on the process. Our contributions range from the theory of data-driven control, such as convergence, identifiability and statistical properties of the estimates, to methods and software that engineers can apply on real plants.

Virtual Reference Feedback Tuning

One-shot controller tuning from a single batch of data. We have extended VRFT to non-minimum phase plants, multivariable and cascade systems, nonlinear controllers and disturbance rejection (Virtual Disturbance Feedback Tuning).

Optimal Controller Identification

Data-driven model reference control formulated as a prediction error identification problem, for single-loop and multivariable systems, including non-minimum phase plants.

Iterative Methods & The H2 Framework

Convergence and robustness of iterative data-based tuning such as IFT and CbT, and the unifying H2 framework presented in our Springer book.

System Identification

Prediction error methods, identifiability and uniqueness of the minimum, input design, regularized estimation and errors-in-variables problems.

Dynamic Networks

Identifiability of dynamic networks with partial node measurements and singular noise spectra, and optimal excitation and measurement patterns.

PID & Resonant Controllers

Tuning rules and the generalized forced oscillation method for PID and proportional-resonant controllers, and data-driven performance monitoring and retuning.

Applications

Power electronics (DC–DC converters and uninterruptible power supplies), power systems (wide-area damping controllers), industrial wireless networks (WirelessHART), multivariable process control on pilot plants, bioreactors and epidemic models.

Experiments on real plants

pyvrft

An open-source Python package for Virtual Reference Feedback Tuning, by E. Boeira and D. Eckhard, published in SoftwareX (2020).

Python · Open source

Data-driven Controller Design: The H2 Approach

A. S. Bazanella, L. Campestrini and D. Eckhard. A comprehensive theoretical treatment of the H2 approach to data-driven control design, with practical designs for thermal, fluid processing and electromechanical processes.

Springer, 2012